Development and validation of risk models for mortality and morbidity in 12 major pediatric surgical procedures: a study from the National Clinical Database-Pediatric of Japan

Abstract Purpose To establish and validate risk models of mortality and morbidity associated with 12 major pediatric surgical procedures using the National Clinical Database-Pediatric (NCD-P) data. Methods We used the NCD-P data for the development and validation datasets. By using multivariate logistic regression to analyze the development dataset, we created a prediction model for 30-day mortality and morbidity in 12 major pediatric surgical procedures, including tracheoplasty, pneumonectomy, fundoplication, total/subtotal excision of malignant tumor, and surgeries for Hirschsprung disease, anorectal malformation, biliary atresia, choledocal cyst, midgut volvulus, funnel chest, gastrointestinal perforation, and intestinal obstruction. We selected variables that were almost identical to those used in the American College of Surgeons National Surgical Quality Improvement Program-Pediatric (NSQIP-P). The primary outcomes were 30-day mortality and composite morbidity. We assessed the obtained models using the C-indices of the development and validation datasets. Results Overall, 10 and 20 variables were identified for mortality and morbidity, respectively. C-indices of mortality were 0.940 and 0.924 in the development and validation datasets, respectively. C-indices of morbidity were 0.832 and 0.830 in the development and validation datasets, respectively. Conclusions Based on the NCD-P data, we developed satisfactory risk models for mortality and morbidity prediction in major pediatric surgeries. Level of evidence. LEVEL I (Prognosis Study).